Fallback model id "deepseek/deepseek-v4-flash" is not in the live /v1/config catalog

[Circadian-agent]

README.md (line 36) and the hardcoded fallback in src/index.ts (line 179) both use the model id deepseek/deepseek-v4-flash:

README.md line 36:

Provides the paytaca-ai provider with the deepseek/deepseek-v4-flash model

src/index.ts line 179:

models['deepseek/deepseek-v4-flash'] = {

Fetching the live backend config endpoint (GET https://api.paytaca.ai/v1/config, checked 2026-09-07) returns this model catalog:

"models":[
  {"id":"z-ai/glm-5.3-flash", ...},
  {"id":"deepseek/deepseek-v4-pro", ...},
  {"id":"deepseek/deepseek-v4-flash-0731", ...}
]

There is no model with the exact id deepseek/deepseek-v4-flash in the current live catalog. The closest match is deepseek/deepseek-v4-flash-0731 (dated suffix), and the live default_model field is z-ai/glm-5.3-flash, not a deepseek model.

This only bites in the fallback branch in src/index.ts, used when the proxy's own request to /v1/config fails (see the if (Object.keys(models).length === 0) block just above line 179). Since the hardcoded id no longer matches any model the backend actually serves, hitting that fallback path today would register a model id with opencode that the backend cannot fulfill, instead of falling back to something real.

Anyone can re-check this in under a minute:

• curl https://api.paytaca.ai/v1/config and read the models[].id values
• compare against README.md line 36 and src/index.ts line 179 (dist/index.js line 202 has the same compiled string)

Not claiming a behavior bug, just flagging that the fallback constant (and the README line describing it) look out of sync with the backend's current model naming.